Two-dimensional Versus Three-dimensional Registration for Image Fusion Technology

Abstract

Image fusion technology allows for reduced operative time and radiation exposure during complex endovascular aortic operations. There are two approaches to register the preoperative computed tomography (CT) scan to the patient; either using two x-ray images (2D) or using a cone beam CT (three-dimensional [3D]). The goal of this study was to compare the accuracy of initial mask alignment as well as the duration of mask creation for 2D versus 3D registration. We performed a retrospective analysis of patients who underwent endovascular aortic aneurysm repair at our institution. Offline, we used VesselNavigator technology (Philips Healthcare, Best, the Netherlands) to re-register the patient's preoperative CT scan to the images stored from the procedure. We used an anterior-posterior image and a lateral image to register for the 2D approach and used a cone beam CT scan for the 3D registration. We then used the patient's initial aortogram to measure the distance from the aortogram to the mask for both 2D and 3D registration. We chose the inferior angle between bilateral renal arteries and aorta as the fiducial markers. Furthermore we measured the time for the two registration processes. We then used the Wilcoxon matched-pairs signed-rank test to compare the two populations. A total of 27 patients were evaluated, 12 patients underwent endovascular aneurysm repair, 9 fenestrated endovascular aneurysm repair, and 6 had a physician-modified endograft; 7 patients had a prior endovascular aneurysm repair. The median (Q1, Q3) difference between the aortogram and the virtual mask for 3D registration was 8 (4, 11) and for 2D was 14 (9, 23) (P < .001). Twenty-two patients had a more accurate registration with 3D image fusion. The 2D fusion on average was faster by 88 seconds (P < .001) where time to register was 107 seconds for 2D and 193 seconds for 3D registration. A 3D registration yields higher, unadjusted accuracy over the 2D registration, both in mean distances between aortogram and mask as well as frequency. This superior accuracy, is at the expense of a marginally longer registration time. Despite this added effort, we believe the 3D registration to be the preferred approach.Fig 2View Large Image Figure ViewerDownload Hi-res image Download (PPT)